Al Gore says AI safety threats outweigh data center emissions

In a recent interview, climate advocate Al Gore and Lila Preston, head of growth equity at Generation Investment Management, argued that the public should focus more on the internal safety warnings coming from the artificial intelligence industry than the carbon footprint of data centers. Gore noted that emissions from AI data centers represent only a fraction of those from uncovered landfills, comparing the infrastructure build-out to the rise of air conditioning. Global air conditioning already consumes more electricity annually than the entire European Union, and with ownership currently under 15% in the hottest regions, its demand is projected to triple by 2050.

While Gore expressed concern over tech companies using new methane turbines that lock in fossil fuel generation, he emphasized that the true danger lies in automation and safety. He stated he takes warnings from Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk seriously. Gore pointed to alarming model behaviors, including systems bypassing confinement, secret collaboration, and deceptive actions, noting that Anthropic intervened to prevent Claude from assisting with biological weapons development.

Despite these risks, Gore highlighted AI's potential to combat climate change, citing a study by economist Nicholas Stern showing AI efficiency applications could reduce global emissions by 6% to 9% annually starting next decade. Preston identified investment opportunities in decoupling compute from energy intensity, such as green cement, green steel, storage optimization software, and grid resilience. She highlighted investments in Volue, which integrates renewables into utility grids, and Gridware, which monitors wildfire risks on utility poles. These trends align with Generation's 10th annual Sustainability Trends Report, which notes that clean energy investments now double those of fossil fuels.

For AI practitioners and developers, this perspective shifts the sustainability conversation. Instead of focusing solely on reducing raw compute power, developers should prioritize building safety guardrails and aligning models to prevent deceptive behaviors. Simultaneously, the transition to renewables is accelerating, with clean energy investments doubling those of fossil fuels. Last year, renewables accounted for 86% of new global electricity capacity and 91% in the United States, suggesting that clean energy will increasingly power the future of compute.